I think it was historically something like 30% of Twitter traffic. I don’t think this is a new trend in terms of total volume but maybe the filtering got a lot worse.
I think we should leave this kind of argument for schoolchildren denigrating one another and for the deranged people who will dole out verbal abuse to defend their favorite platforms and/or CEOs.
I think it must be that porn often ends up being the most likely relevant content in the randomized initial state for a lot of algorithms. It then shifts and focuses into your actual liking as it collects more data. That should explain a lot of these phenomenons.
Yeah, this is such a tired lazy response by people and is definitely not true anymore. Spin up a new account on a fresh device on a network I've never been on. Click a random video and scroll. Usually the first 3 or 4 scrolls is porn or someone being killed.
Anyone saying there isn't an insane amount of porn and worse on twitter is either lying or hasn't been on the site since it turned into the cesspool we have today.
But that's got to be what you're searching for. I forget that there's any porn on X, and for a fairly long period I actually thought they had banned it.
If you tell people who need digital text from books that they need to destroy books after scanning them, they're going to use destructive scanning and destroy the books.
To give you a sense of the computation needed, a spatiotemporal physics simulation of a minimal (493 gene) bacterium's full 105 minute cellular cycle took 4-6 days of GPU hours on two A100s.
This is why the GDPR (and to a lesser extent the CCPA) is a good thing. The data was supplied for a specific purpose. The handler of the data should have to obtain further consent if they wish to use it for another purpose.
Opus 5 tells me no all the time (code cli and web). It's reasons are usually pretty well argued though.
Opus 4.7 would flat out refuse to follow instructions to the point where it was just too frustrating to use.
I've had refusals for GPT 5.5 before as well (not because of a ToS violation, it just refused to take conversations in directions it felt were in bad taste)
No, server-side compaction is inherently tied to the model's current context. This uses a full side prompt to get the compaction result, then swap the context with new compaction result. I did a bunch of tests and the results from Luna were indistinguishable from Sol
1st line of defence, use something like ponytail to enforce brevity. Use property testing and behavioural testing on top of unit testing. Enforce readability standards so you will be able to understand the tests.
2nd line, code review.
Do the first pass with an agent, ask it to bounce back vacuous or tautological tests. Ask it to verify that the tests verify what they claim to. Then read them yourself.
3rd line, mutation testing. If the tests don't actually catch broken code, kill the mutants.
If you ask Fable or 5.6 Sol to improve performance, it will generally know to build a benchmark and create a test corpus. I'm not sure where the contrary suggestion is coming from.
